Why Is Remote Sensing of Amazon Forest Greenness So ...

Why Is Remote Sensing of AmazonForest Greenness So Challenging?Arindam Samanta*

Department of Geography and Environment, Boston University, Boston, and Atmosphericand Environmental Research Inc., Lexington, Massachusetts

Sangram Ganguly

BAERI/NASA Ames Research Center, Moffett Field, California

Eric Vermote

Department of Geography, University of Maryland, College Park, College Park, Maryland

Ramakrishna R. Nemani

Biospheric Science Branch, NASA Ames Research Center, Moffett Field, California

Ranga B. Myneni

Department of Geography and Environment, Boston University, Boston, Massachusetts

Received 23 January 2012; accepted 24 March 2012

ABSTRACT: The prevalence of clouds and aerosols and their impact onsatellite-measured greenness levels of forests in southern and central Amazoniaare explored in this article using 10 years of NASA Moderate ResolutionImaging Spectroradiometer (MODIS) greenness data: normalized differencevegetation index (NDVI) and enhanced vegetation index (EVI). During the wetseason (October–March), cloud contamination of greenness data is pervasive;

* Corresponding author address: Arindam Samanta, Atmospheric and Environmental Re-search Inc., 131 Hartwell Avenue, Lexington, MA 02421.

E-mail address: [email protected]

Earth Interactions d Volume 16 (2012) d Paper No. 7 d

DOI: 10.1175/2012EI440.1

Copyright � 2012, Paper 16-007; 34019 words, 5 Figures, 0 Animations, 0 Tables.http://EarthInteractions.org

nearly the entire region lacks uncorrupted observations. Even in the dry season(July–September), nearly 60%–66% of greenness data are corrupted, mainlybecause of biomass burning aerosol contamination. Under these conditions,spectrally varying residual atmospheric effects in surface reflectance data in-troduce artifacts into greenness indices; NDVI is known to artificially decrease,whereas EVI, given its formulation and use of blue channel surface reflectancedata, shows artificial enhancement, which manifests as large patches of en-hanced greenness. These issues render remote sensing of Amazon forest green-ness a challenging task.

KEYWORDS: Amazon forests; Greenness; Vegetation index; Remote sensing

1. IntroductionVegetation indices (VIs) are radiometric measures of photosynthetically active

radiation absorbed by canopy chlorophyll and are therefore good surrogate mea-sures of the physiologically functioning surface greenness level in a region(Myneni et al. 1995). The normalized difference vegetation index (Tucker 1979)of the National Oceanic and Atmospheric Administration (NOAA) Advanced VeryHigh Resolution Radiometer (AVHRR) has been used effectively to analyzegreenness changes at seasonal (Ferreira and Huete 2004) to interannual time scales(Asner et al. 2000; Gurgel and Ferreira 2003; Dessay et al. 2004) in the Amazon.The shortcomings of the AVHRR NDVI data—calibration degradation, loss oforbit, lack of satisfactory atmospheric correction, broad spectral bands, saturationin dense vegetation, insufficient validation, etc.—are for the most part remediedin the National Aeronautics and Space Administration (NASA) Moderate Reso-lution Imaging Spectroradiometer (MODIS) enhanced vegetation index (EVI) data(Justice et al. 1998; Vermote et al. 2002; Huete et al. 2002). Algorithm refinementsthrough feedback from concerted validation efforts and multiple reprocessings ofthe growing MODIS data archive have led to progressively improved research-quality products ideally suited for studies of greenness dynamics in complex ec-osystems such as the Amazon (e.g., Huete et al. 2006; Xiao et al. 2006; Myneniet al. 2007; Samanta et al. 2011a; Samanta et al. 2012; Xu et al. 2011).

Monitoring greenness of Amazon forests with satellite measurements is fraughtwith several pitfalls, the major being corruption from atmospheric effects leadingto artificial changes in vegetation indices that are unrelated to any real changes onthe ground. For example, Saleska et al. (Saleska et al. 2007) observed a greeninganomaly in response to the 2005 drought, but this has been shown to be an artifactof atmospheric corruption of EVI and not observed in a later version of the samedata (Samanta et al. 2010; Samanta et al. 2011b). This article is thus motivated bythe need to comprehensively assess cloud and aerosol prevalence and corruption ofgreenness data, using a decade’s worth of MODIS VI data.

2. Data

2.1. Greenness data: VI

These are satellite data-based measurements of vegetation greenness producedby NASA using blue [blue bidirectional reflectance factor (BRFBLUE); 459–479 nm(nm)], red (BRFRED; 620–670 nm) and near-infrared (NIR) (BRFNIR; 842–876 nm)


band surface reflectance data from the MODIS instrument aboard the Terra andAqua satellites (USGS 2011a; Huete et al. 2002). VIs consist of NDVI and EVI.NDVI is a radiometric measure of photosynthetically active radiation absorbed bycanopy chlorophyll and therefore is a good surrogate measure of the physiologi-cally functioning surface greenness level in a region (Myneni et al. 1995). NDVIhas been used in many studies of vegetation dynamics in the Amazon (e.g., Asneret al. 2000; Dessay et al. 2004; Ferreira and Huete 2004). EVI is also a measure ofgreenness that generally correlates well with ground measurements of photosyn-thesis (e.g., Rahman et al. 2005; Sims et al. 2008) and found to be especially usefulin high biomass tropical broadleaf forests like the Amazon (Huete et al. 2006). Weused versions—also called collections—4 and 5 of Terra MODIS EVI data. Col-lection 5 (C5) is the latest version, superseding all previous versions.

The vegetation indices 16-day L3 global 1 km (MOD13A2) dataset contains VIat 1 km 3 1 km spatial resolution and 16-day frequency. This 16-day frequencyarises from compositing: that is, assigning one best-quality EVI value to representa 16-day period. This dataset is available in tiles (108 3 108 at the equator) ofsinusoidal projection; 16 such tiles cover the Amazon region (approximately108N–208S, 808–458W). The data were obtained from the NASA Land ProcessesData Active Archive Center (LP DAAC) (USGS 2011a) for the period February2000–December 2009. A previous version of EVI data, collection 4 (C4), was alsoused in this study, although this dataset is now decommissioned and deleted fromthe NASA LP DAAC archives [the July–September (JAS) C4 Terra MODIS EVIdata but not the corresponding quality flags for the years 2000–05 were obtainedfrom Saleska et al. (2007)]. We have also used two other datasets. First, we used thevegetation indices monthly L3 global 0.05deg CMG (MOD13C2) dataset, whichcontains EVI at 0.058 3 0.058 spatial resolution and a monthly frequency. Second,we used the vegetation indices 16-day L3 global 0.05deg CMG (MOD13C1)dataset, which contains EVI at 0.058 3 0.058 spatial resolution and a 16-dayfrequency. These are cloud-free spatial composites of MOD13A2 (USGS2011b). These were obtained from the NASA LP DAAC (USGS 2011b) for theperiod February 2000–December 2009. We also were able to obtain C4MOD13C1 data with quality flags for 2004–06 (USGS 2011b).

2.2. Land-cover data

Land-cover information was obtained from the MODIS Terra land cover typeyearly L3 global 1 km SIN grid product (MOD12Q1). This is the official NASA C5land-cover dataset (USGS 2011c; Friedl et al. 2010). It consists of five land-coverclassification schemes at 0.5 km 3 0.5 km spatial resolution. The InternationalGeosphere Biosphere Programme (IGBP) land-cover classification scheme wasused to identify forest pixels in the Amazon region.

2.3. Aerosol data

Daily aerosol optical thickness (AOT) was obtained from the level 3 daily jointaerosol/water vapor/cloud product (MOD08_D3) at 18 3 18 spatial resolu-tion. These were obtained from the NASA Level 1 and Atmosphere Archive and


Distribution System (LAADS) (NASA 2011a) for the months of July–Septemberof the years 2000–06. We used the optical_depth_land_and_ocean_mean_meandata field from this product, which contains AOT at 550 nm. Similarly, monthlyAOT was obtained from the level 3 monthly joint aerosol/water vapor/cloudproduct (MOD08_M3) at 18 3 18 spatial resolution. We used the corrected_optical_depth_land_QA_mean data field from this product, which contains AOTat 550 nm.

3. MethodsVI products are generated from surface reflectance data that have undergone

atmospheric correction. However, residual atmospheric effects that remain in thesedata can be assessed by using 16-bit quality flags in both 1 km 3 1 km as well asthe 0.058 3 0.058 VI products. Sets of bits, from these 16 bits, are assigned to flagspertaining to clouds and aerosols, as well as flags that provide aggregate measuresof data quality called VI usefulness indices, as detailed below.

Cloud quality flags are single bit (binary) flags indicating the presence (1) orabsence (0) of clouds. In MOD13A2, each 1 km 3 1 km 16-day VI compositevalue is accompanied by three binary cloud quality flags: adjacent cloud detected(bit 8), mixed clouds (bit 10), and possible shadows (bit 15). In the case ofMOD13C1/C2, each 0.058 3 0.058 monthly EVI value has two binary cloud qualityflags: adjacent cloud detected (bit 8) and mixed clouds (bit 10).

Aerosol quality flag, 2 bits in precision, provides information on aerosol contentand is named aerosol quantity. It occupies the same bit positions, bits 6 through 7,in both MOD13A2 and MOD13C1/C2. The aerosol quality flag can assume one offour values: climatology (00), low (01), average (10), and high (11). The low,average, and high values refer to aerosol optical depth (AOT) at 550 nm less than0.2, between 0.2 and 0.5, and greater than 0.5, respectively (Vermote and Vermuelen1999). On the other hand, the climatology value indicates that the actual AOT isunknown, most likely because of presence of clouds, and climatological (long-termaverage) AOT is used in the process of atmospheric correction (Vermote andVermuelen 1999).

The VI usefulness flag, 4 bits in precision, provides an aggregate measure ofEVI quality. It occupies the same bit positions, bits 2 through 5, in both MOD13A2and MOD13C1/C2 and can assume values from 0 (0000: best quality) to 15 (1111:not useful).

4. Results and discussionRemote sensing of the Amazon forests at solar wavelengths is complicated due

to persistent cloud and aerosols presence. Atmospheric corruption of surface re-flectance data, from which the vegetation indices are evaluated, depends on pre-vailing cloud and aerosol optical thicknesses, among other factors (Vermote andKotchenova 2008; Tanre et al. 1992). The two data quality indicators, quality flagsand VI usefulness indices, accompanying the VI data indicate extensive cloud andaerosol corruption over the entire Amazon region (Figures 1a,b); about 81% of theforest pixels have at least one of the six 16-day VI composites in the third quarter of2005 corrupted with clouds or aerosols and 56% have two or more composites


Figure 1. Spatial patterns of atmosphere corruption of C5 VI (1 km 3 1 km) dataduring the JAS quarter. (a) Number of 16-day VI composites in the JASquarter of 2005 with quality flags indicating cloud (adjacent cloud, mixedclouds, and possible shadow) or aerosol (climatology and high aerosols)corruption. (b) VI usefulness index values corresponding to (a). When a VIvalue is atmosphere corrupted only once, the corresponding VI useful-ness index is its VI usefulness index value. When a pixel is atmospherecorrupted twice or more, the corresponding index is the mode of the VIusefulness index values. When a pixel is not atmosphere corrupted evenonce, the corresponding index is the mode of all six VI usefulness indexvalues. (c) Average number of 16-day VI composites in the JAS quarter of2000–06 (excluding 2005) with quality flags indicating clouds (adjacentcloud, mixed clouds, and possible shadow). (d) As in (c), but with qualityflags indicating aerosols (climatology and high aerosols). The ellipse ineach panel denotes the dominant drought-affected region during JAS2005.


similarly corrupted. Cloud and aerosol presence and their radiation scatteringproperties are highly dynamic, both spatially and temporally; nevertheless, it isinstructive to look at some summary statistics relevant to this study.

Clouds are especially pervasive north of the equator during the July–Septemberquarter; here, on average, 90% of the pixels have two or more of the six 16-day dryseason VI composites corrupted with clouds or their shadows (Figure 1c). Thiscorruption is only somewhat less, 75%, in 2005 (figure not shown for brevity).South of the equator, cloud contamination is much less extensive—on averageabout 22%—because it is dry season here and even lesser during 2005 because ofthe drought (9%). Within the drought-stricken region itself, not surprisingly, onlyabout 6% of the pixels exhibit cloud contamination during the third quarter of2005.

Aerosol presence is more extensive in the Amazon during the dry season be-cause of biomass burning (Eck et al. 1998; Schafer et al. 2002a; Schafer et al.2002b; Artaxo et al. 1998; Artaxo et al. 2002). About 74% of the pixels north of theequator had two or more of the six 16-day VI composites during July–September2005 corrupted due to failure of the aerosol retrieval algorithm because of cloud-iness and subsequent use of climatological aerosol amount for atmospheric cor-rection (Figure 1d); the average number is even higher, 86%, based on data fromother years (2000–06; see Figure S2 of Samanta et al. 2010). South of the equatorin the drought-stricken region, the respective numbers are 36% for 2005 and 19%for other years, but the corruption here is due to high aerosol amount, especiallyduring the months of August and September (Samanta et al. 2010). These numbersmay appear low, but that is misleading, because different pixels show aerosol cor-ruption at different times and, as a result, few pixels within the study region actuallyhave the full complement of uncorrupted time series data required for an unbiasedanalysis.

The spatiotemporal evolution of aerosol amount during the 2005 dry season isapparent from Figure 2: aerosol optical depths in excess of 0.5 at 550 nm begin toappear during the first half of August in the southern Amazon, covering nearly theentire drought-stricken region, and over the next 45 days the aerosol loads growrapidly, reaching values of 1 and above, which are anomalously high when com-pared to other years during the period 2000–06, as noted by Koren et al. (Korenet al. 2007) and Bevan et al. (Bevan et al. 2009). In the drought-affected region,82%, 58%, and 84% of the pixels have AOT values greater than 0.5 during the lastthree 16-day compositing periods in the third quarter of 2005 (Figures 2d–f);conservatively speaking, over 80% of the drought-affected region was under a thickhaze for 40% of the time during the third quarter of 2005. This threshold AOTvalue of 0.5 at 550 nm is generally taken as the limit beyond which the corre-sponding surface reflectance data, hence VI, are deemed not useful for remotesensing of vegetation (NASA 2011b; MODIS 2011; Didan and Huete 2006;Vermote and Kotchenova 2008). The fact that the spatiotemporal patterns of AOTmagnitudes change from year to year (Koren et al. 2007; Bevan et al. 2009)invalidates any arguments regarding the validity of differences in corrupted EVIvalues between years as indicative of actual greenness changes (as in Saleskaet al. 2007).

It is instructive to contrast the actual day-to-day variation in AOT over thedrought-affected region during 2005 with other years, as in Figure 3. Starting from


about day 216 (early August), AOT remains much greater than 0.5 throughoutthe entire 2005 dry season, rendering accurate atmospheric correction of MODISmeasurements impossible. Compared to other years, the 2005 dry season wasindeed extremely hazy, due in large part to a very high number of biomass firesreported (Aragao et al. 2007). By examining the date stamp of each pixel in thedrought-affected region, we find that nearly 100% of the VI values during thesecond half of August and most of September were aerosol corrupted (Figure 3).This renders any analysis of VI data in this region at 16-day scale fraught withbiases. Even at monthly scale, we find that nearly 60%–66% of the drought-stricken region has insufficient valid VI data, which makes it difficult to draw anyconclusions on the greenness dynamics of Amazon forests because of cloud andaerosol corruption.

Atmospheric corruption of is known to artificially reduce NDVI values (e.g.,Holben 1986; Huete et al. 2002). The EVI was designed to improve upon NDVIby introducing aerosol resistance (Huete et al. 2002): that is, mitigate aerosoleffects by using the blue band to correct for aerosol influence in the red band(Kaufman and Tanre 1992). This raises an important question: how do residualaerosol effects (given that the input surface reflectance data were generated after

Figure 2. Spatial patterns of MODIS AOT at 550 nm, at 18 3 18 spatial resolution,during the dry season (JAS) of 2005. Shown here are patterns of mean AOTover the 16-day period corresponding to each of the six 16-day EVIcomposites covering JAS 2005. Negative AOT values are colored black.The ellipse in each panel denotes the dominant drought-affected regionduring JAS 2005.


atmospheric correction) impact magnitudes of C4 and C5 EVI data? To identifycorrupted and uncorrupted pixels, we used C5 data quality filters, which super-seded C4 filters, for both C4 and C5 EVI data (cf. section 3). This allowed anaccurate assessment of atmosphere-corruption effects on EVI across collections bynot introducing errors related to changes in data filters between collections. This isfurther justified by a high degree of similarity in cloud and aerosol flags betweenC4 and C5 data reported in Samanta et al. (Samanta et al. 2011b). The difference inuncorrupted EVI values between C4 and C5 is, as to be expected, negligible (0.02or less compared to an average EVI value of 0.5), both during the dry season of2005 (Figure 4a) and in other years as well (results not shown for brevity). In-terestingly, the difference between corrupted and uncorrupted C5 EVI values issmall (Figure 4b), which must be due to refinements to atmospheric correction(Vermote and Kotchenova 2008) and EVI algorithms (Didan and Huete 2006),because the difference between these two classes of pixels in C4 is quite large:corrupted EVI values are greater than uncorrupted values by 0.1 and more, espe-cially in areas of high aerosol amounts (cf. Figures 1d, 4c). Aside from the largedifference, the sign of this change is counterintuitive, because atmospheric cor-ruption decreases estimates of vegetation greenness (Holben 1986; van Leeuwenet al. 2006), which is the reason for maximum-value compositing of vegetationindex data: that is, selecting the largest vegetation index value among a set of dailyvalues to represent a compositing period (16 days in the case of MODIS EVI). Thissuggests unusual behavior of the C4 EVI algorithm in the face of residual

Figure 3. Distribution (percent of pixels) of daily C5 VI (1 km 3 1 km) retrievals inintact forests in the drought-affected area south of the equator for all six16-day VI composites during JAS 2005 (left axis). Daily MODIS AOT aver-aged over the same region (right axis) for JAS 2005 (in red) and 2000–06(excluding 2005, in black).


Figure 4. Spatial patterns of EVI differences during the 2005 dry season (JAS) at1 km 3 1 km spatial resolution. (a) Difference between uncorrupted C4and C5 EVI values. (b) Difference between corrupted and uncorrupted C5EVI values. (c) Difference between corrupted and uncorrupted C4 EVIvalues. (d) As in (c), but using C4 flags (0.058 3 0.058 spatial resolution). A16-day composite EVI value is deemed corrupt if the aerosol-quantityquality flag indicates high aerosols. Likewise, a 16-day composite EVIvalue is deemed uncorrupted if the aerosol-quantity quality flag does notindicate high aerosols. In all cases, cloud contamination (mixed clouds,adjacent clouds, and shadows) is absent. The ellipse in the panels showsthe drought-affected region.


atmospheric effects in surface reflectance data, from which the EVI is evaluated buthas been fixed in the C5 algorithm. This behavior, as explained below, would resultin artificially enhanced greening of vegetation when atmosphere-corrupted data areincluded in the analysis.

When surface reflectance data have (residual) aerosol corruption, as in C4 data,the resulting EVI values tend to be artificially inflated. This artifact can be traced tothe EVI formulation; at a given near-infrared reflectance value, EVI decreaseswhen red and blue reflectance values increase simultaneously, which correctly de-notes increased obscuration of the surface under thicker atmospheres (Figure 5a).On the other hand, if only blue reflectance value increases [i.e., the ratio of blue tored reflectance (at given red reflectance) increases], EVI increases dramatically(Figure 5). Atmospheric effect at blue is much greater than at red, because scat-tering of solar radiation in the atmosphere increases nonlinearly with decrease inwavelength; this can easily result in blue band reflectance being greater than redband reflectance in situations of high aerosol loads, which will result in spuriouslyhigh values of EVI. Moreover, EVI will also increase if near-infrared reflectanceincreases as a result of incomplete atmospheric correction. These effects are dem-onstrated very well in C4 reflectance data (Figures 5c,d). Besides, inconsistencybetween multisensor EVI arising from problematic blue band reflectance valueswas noted in Fensholt et al. (Fensholt et al. 2006). The fact that such large dif-ferences between corrupted and uncorrupted EVI values are not seen in C5 (cf.Figures 4c,d vs. Figure 4b) could be due to better estimation of aerosol amount,thus affording more accurate atmospheric correction of MODIS measurements(Vermote and Kotchenova 2008), and to C5 EVI algorithm refinements (Didan andHuete 2006).

5. ConclusionsCloud and aerosol corruption of satellite-based estimates of vegetation green-

ness is pervasive in the Amazon; nearly the entire region lacks a sufficient numberof uncorrupted observations during the months of October–March. Even in the dryseason (e.g., July–September 2005), about 81% of the forest pixels have at leastone of the six 16-day VI composites corrupted with clouds or aerosols and 56%have two or more composites similarly corrupted. Under these conditions, spec-trally varying residual atmospheric effects in surface reflectance data introduceartifacts into greenness indices; the EVI is particularly susceptible to such artifactsgiven its formulation and use of blue channel surface reflectance data. Such arti-facts due to aerosol corruption—large patches of enhanced greenness—are espe-cially obvious in collection 4 EVI data.

This analysis does not imply that satellite-measured greenness data are com-pletely ineffective for studying vegetation changes in the Amazon but rather dem-onstrates that there is a need to rigorously evaluate these greenness data, asappropriate to a given application. For instance, future research should investigatehow such greenness data could be validated against, as well as used synergisticallywith, data from other sources. These sources include other satellite sensors withdifferent spatial, temporal, and spectral band characteristics (multispectral, hyper-spectral, radar, and microwave) and, importantly, data from field measurements.Further, the impact of future improvements in atmospheric correction algorithms


Figure 5. Sensitivity of EVI to blue band reflectance. (a) Theoretical calculationsshowing changes in EVI when only blue band reflectance increases (red)and when both blue and red band reflectances increase (green). NIRband reflectance is fixed at 0.35 in these calculations. (b) C5 (0.058 30.058) EVI vs blue to red reflectance ratio in Amazon forests during JAS2005. The blue to red ratios shown here are estimated from blue bandreflectance values varying from 0.029 to 0.084 and a narrow interval ofred band reflectance (0.06–0.07). NIR reflectance varied between 0.34and 0.36. (c) Spatial patterns of C4 NIR reflectance differences betweencorrupted and uncorrupted, using C4 flags, during the 2005 dry season(JAS) at 0.058 3 0.058 spatial resolution. (d) As in (c), but for blue to redreflectance ratio.


(e.g., collection 6 and beyond) on the effectiveness of greenness data should bestudied. Finally, there is a need to understand what other regions in the world showsimilar issues.

Acknowledgments. This research was funded by the NASA Earth Science Enterprise.

References

Aragao, L., Y. Malhi, R. M. Roman-Cuesta, S. Saatchi, L. O. Anderson, and Y. E. Shimabukuro,2007: Spatial patterns and fire response of recent Amazonian droughts. Geophys. Res. Lett.,34, L07701, doi:10.1029/2006GL028946.

Artaxo, P., E. T. Fernandes, J. V. Martins, M. A. Yamasoe, P. V. Hobbs, W. Maenhaut, K. M. Longo,and A. Castanho, 1998: Large-scale aerosol source apportionment in Amazonia. J. Geophys.Res., 103 (D24), 31 837–31 847.

——, and Coauthors, 2002: Physical and chemical properties of aerosols in the wet and dry seasonsin Rondonia, Amazonia. J. Geophys. Res., 107, 8081, doi:10.1029/2001JD000666.

Asner, G. P., A. R. Townsend, and B. H. Braswell, 2000: Satellite observation of El Nino effects onAmazon forest phenology and productivity. Geophys. Res. Lett., 27, 981–984.

Bevan, S. L., P. R. J. North, W. M. F. Grey, S. O. Los, and S. E. Plummer, 2009: Impact of atmosphericaerosol from biomass burning on Amazon dry-season drought. J. Geophys. Res., 114, D09204,doi:10.1029/2008JD011112.

Dessay, N., H. Laurent, L. A. T. Machado, Y. E. Shimabukuro, G. T. Batista, A. Diedhiou, andJ. Ronchail, 2004: Comparative study of the 1982-1983 and 1997-1998 El Nino events overdifferent types of vegetation in South America. Int. J. Remote Sens., 25, 4063–4077.

Didan, K., and A. R. Huete, 2006: MODIS vegetation index product series collection 5 changesummary. 17 pp. [Available online at http://landweb.nascom.nasa.gov/QA_WWW/forPage/MOD13_VI_C5_Changes_Document_06_28_06.pdf.]

Eck, T. F., B. N. Holben, I. Slutsker, and A. Setzer, 1998: Measurements of irradiance attenuationand estimation of aerosol single scattering albedo for biomass burning aerosols in Amazonia.J. Geophys. Res., 103 (D24), 31 865–31 878.

Fensholt, R., I. Sandholt, and S. Stisen, 2006: Evaluating MODIS, MERIS, and VEGETATION—Vegetation indices using in situ measurements in a semiarid environment. IEEE Trans. Geosci.Remote Sens., 44, 1774–1786.

Ferreira, L. G., and A. R. Huete, 2004: Assessing the seasonal dynamics of the Brazilian Cerradovegetation through the use of spectral vegetation indices. Int. J. Remote Sens., 25, 1837–1860.

Friedl, M. A., D. Sulla-Menashe, B. Tan, A. Schneider, N. Ramankutty, A. Sibley, and X. M. Huang,2010: MODIS collection 5 global land cover: Algorithm refinements and characterization ofnew datasets. Remote Sens. Environ., 114, 168–182.

Gurgel, H. C., and N. J. Ferreira, 2003: Annual and interannual variability of NDVI in Brazil and itsconnections with climate. Int. J. Remote Sens., 24, 3595–3609.

Holben, B. N., 1986: Characteristics of maximum-value composite images from temporal AVHRRdata. Int. J. Remote Sens., 7, 1417–1434.

Huete, A., K. Didan, T. Miura, E. P. Rodriguez, X. Gao, and L. G. Ferreira, 2002: Overview of theradiometric and biophysical performance of the MODIS vegetation indices. Remote Sens.Environ., 83 (1–2), 195–213.

——, and Coauthors, 2006: Amazon rainforests green-up with sunlight in dry season. Geophys.Res. Lett., 33, L06405, doi:10.1029/2005GL025583.

Justice, C. O., and Coauthors, 1998: The Moderate Resolution Imaging Spectroradiometer(MODIS): Land remote sensing for global change research. IEEE Trans. Geosci. RemoteSens., 36, 1228–1249.


Kaufman, Y. J., and D. Tanre, 1992: Atmospherically resistant vegetation index (ARVI) for EOS-MODIS. IEEE Trans. Geosci. Remote Sens., 30, 261–270.

Koren, I., L. A. Remer, and K. Longo, 2007: Reversal of trend of biomass burning in the Amazon.Geophys. Res. Lett., 34, L20404, doi:10.1029/2007GL031530.

MODIS, cited 2011: Validation status. MODIS Land Surface Reflectance Science ComputingFacility. [Available online at http://modis-sr.ltdri.org/validation/accuracy_uncertainty.html.]

Myneni, R. B., F. G. Hall, P. J. Sellers, and A. L. Marshak, 1995: The interpretation of spectralvegetation indexes. IEEE Trans. Geosci. Remote Sens., 33, 481–486.

——, and Coauthors, 2007: Large seasonal swings in leaf area of Amazon rainforests. Proc. Natl.Acad. Sci. USA, 104, 4820–4823.

NASA, cited 2011a: LAADS web. NASA Level 1 and Atmosphere Archive and DistributionSystem. [Available online at http://ladsweb.nascom.nasa.gov/.]

——, cited 2011b: Product quality documentation for MOD13A2, C5. [Available online at http://landweb.nascom.nasa.gov/cgi-bin/QA_WWW/detailInfo.cgi?prod_id5MOD13A2&ver5C5.]

Rahman, A. F., D. A. Sims, V. D. Cordova, and B. Z. El-Masri, 2005: Potential of MODIS EVI andsurface temperature for directly estimating per-pixel ecosystem C fluxes. Geophys. Res. Lett.,32, L19404, doi:10.1029/2005GL024127.

Saleska, S. R., K. Didan, A. R. Huete, and H. R. da Rocha, 2007: Amazon forests green-up during2005 drought. Science, 318, 612–612.

Samanta, A., S. Ganguly, H. Hashimoto, S. Devadiga, E. Vermote, Y. Knyazikhin, R. R. Nemani,and R. B. Myneni, 2010: Amazon forests did not green-up during the 2005 drought. Geophys.Res. Lett., 37, L05401, doi:10.1029/2009GL042154.

——, M. H. Costa, E. L. Nunes, S. L. Vieira, L. Xu, and R. Myneni, 2011a: Comment on ‘‘Drought-induced reduction in global terrestrial net primary production from 2000 through 2009.’’Science, 333, 1093, doi:10.1126/science.1199048.

——, S. Ganguly, and R. B. Myneni, 2011b: MODIS enhanced vegetation index data do not showgreening of Amazon forests during the 2005 drought. New Phytol., 189, 12–15.

——, and Coauthors, 2012: Seasonal changes in leaf area of Amazon forests from leaf flushing andabscission. J. Geophys. Res., G01015, doi:10.1029/2011JG001818.

Schafer, J. S., B. N. Holben, T. F. Eck, M. A. Yamasoe, and P. Artaxo, 2002a: Atmospheric effectson insolation in the Brazilian Amazon: Observed modification of solar radiation by cloudsand smoke and derived single scattering albedo of fire aerosols. J. Geophys. Res., 107, 8074,doi:10.1029/2001JD000428.

——, T. F. Eck, B. N. Holben, P. Artaxo, M. A. Yamasoe, and A. S. Procopio, 2002b: Observedreductions of total solar irradiance by biomass-burning aerosols in the Brazilian Amazon andZambian Savanna. Geophys. Res. Lett., 29, 1823, doi:10.1029/2001GL014309.

Sims, D. A., and Coauthors, 2008: A new model of gross primary productivity for North Americanecosystems based solely on the enhanced vegetation index and land surface temperature fromMODIS. Remote Sens. Environ., 112, 1633–1646.

Tanre, D., B. N. Holben, and Y. J. Kaufman, 1992: Atmospheric correction algorithm for NOAA-AVHRR products—Theory and application. IEEE Trans. Geosci. Remote Sens., 30, 231–248.

Tucker, C. J., 1979: Red and photographic infrared linear combinations for monitoring vegetation.Remote Sens. Environ., 8, 127–150.

USGS, cited 2011a: Vegetation indices 16-day L3 global 1km. NASA Land Processes Data ActiveArchive Center, MOD13A2. [Available online at https://lpdaac.usgs.gov/lpdaac/products/modis_products_table/vegetation_indices/16_day_l3_global_1km/mod13a2.]

——, cited 2011b: Vegetation indices monthly L3 global 0.05deg CMG. NASA Land ProcessesData Active Archive Center, MOD13C2. [Available online at https://lpdaac.usgs.gov/lpdaac/products/modis_products_table/vegetation_indices/monthly_l3_global_0_05deg_cmg/mod13c2.]

——, cited 2011c: Land cover type yearly L3 global 1km SIN grid. NASA Land Processes DataActive Archive Center, MOD12Q1. [Available online at https://lpdaac.usgs.gov/lpdaac/products/modis_products_table/land_cover/yearly_l3_global_1km2/mod12q1.]


van Leeuwen, W. J. D., B. J. Orr, S. E. Marsh, and S. M. Herrmann, 2006: Multi-sensor NDVI datacontinuity: Uncertainties and implications for vegetation monitoring applications. RemoteSens. Environ., 100, 67–81.

Vermote, E. F., and A. Vermuelen, 1999: Atmospheric correction algorithm: spectral reflectances(MOD09). MODIS Algorithm Theoretical Basis Document, 107 pp. [Available online at http://modis.gsfc.nasa.gov/data/atbd/atbd_mod08.pdf.]

——, and S. Kotchenova, 2008: Atmospheric correction for the monitoring of land surfaces.J. Geophys. Res., 113, D23S90, doi:10.1029/2007JD009662.

——, N. Z. El Saleous, and C. O. Justice, 2002: Atmospheric correction of MODIS data in thevisible to middle infrared: First results. Remote Sens. Environ., 83 (1–2), 97–111.

Xiao, X. M., S. Hagen, Q. Y. Zhang, M. Keller, and B. Moore, 2006: Detecting leaf phenologyof seasonally moist tropical forests in South America with multi-temporal MODIS images.Remote Sens. Environ., 103, 465–473.

Xu, L., A. Samanta, M. H. Costa, S. Ganguly, R. R. Nemani, and R. B. Myneni, 2011: Widespreaddecline in greenness of Amazonian vegetation due to the 2010 drought. Geophys. Res. Lett.,L07402, doi:10.1029/2011GL046824.

Earth Interactions is published jointly by the American Meteorological Society, the American Geophysical

Union, and the Association of American Geographers. Permission to use figures, tables, and brief excerpts

from this journal in scientific and educational works is hereby granted provided that the source is

acknowledged. Any use of material in this journal that is determined to be ‘‘fair use’’ under Section 107 or that

satisfies the conditions specified in Section 108 of the U.S. Copyright Law (17 USC, as revised by P.IL. 94-

553) does not require the publishers’ permission. For permission for any other from of copying, contact one of

the copublishing societies.


Why Is Remote Sensing of Amazon Forest Greenness So ...

Documents

Transcript of Why Is Remote Sensing of Amazon Forest Greenness So ...